[9fans] 9atom

[erik quanstrom]

i have a test image of a new and hopefully improved 9atom @
http://www.quanstro.net/other/9atom2.iso.bz2.
it's not perfect, it's not polished, but it should contain the
basics, be go compatable, contain a copy of nix, be self-compiling,
and a few other things that i'll announce once i have a bit more
confidence that i haven't jobbed the cd.

- erik

[Bence Fábián]

http://ftp.quanstro.net/other/9atom2.iso.bz2

you mean ;)

2013/1/3 erik quanstrom <quan...@quanstro.net>

[faif]

Is that targeting Intel Atom CPUs? I have an EEEPC sitting so that's a good chance to start using Plan9 in a real environment (non VM). The fact that go is provided is tempting :)

[erik quanstrom]

> Is that targeting Intel Atom CPUs? I have an EEEPC sitting so that's
> a good chance to start using Plan9 in a real environment (non VM).
> The fact that go is provided is tempting :)

originally, the goal was to get atom machines working. that goal
was largely met, so the name is somewhat of an anachronism.
it's the same set of kernels i run everwhere.

i don't know anything about any model of the eeepc. ymmv.

there's no go included in 9atom yet, but you should be able
to install it with no problems.

- erik

[John Floren]

If you check out the Go source and don't find any build framework for
Plan 9, you may need to check out the current repo tip to get it.
That's what I did when I installed Go, but it was months ago so YMMV.

john

[Richard Miller]

> originally, the goal was to get atom machines working.

It should be pointed out that the standard Plan 9 distribution
also works on atom processors. It's not the atom cpu itself,
but other details of chipset and peripheral interfaces, which
determines which kernel variant will run on a given motherboard.

[Aram Hăvărneanu]

Downloading from quanstro.net is extremely slow and I was never able
to fully complete the download.

--
Aram Hăvărneanu

[Rox 64]

Hu guys.
Will Plan 9/9atom/9front run on a EEE 1005HA?

[John Floren]

On Fri, Jan 4, 2013 at 8:39 AM, Rox 64 <mro...@gmail.com> wrote:
> Hu guys.
> Will Plan 9/9atom/9front run on a EEE 1005HA?

I have that same netbook. One of your big challenges will probably be
getting the CDROM to boot in a machine without a CDROM drive. There
exist bootable images for USB sticks which contain a full fossil
filesystem; you can start the installation from there. In fact, I had
made a 32-bit bootable USB stick for Nix about a year ago and included
a script to start the installation, I'll see if I can find it.


john

[John Floren]

Following up on this, it looks like the scripts I used are still in
the NIX distribution.

/usr/glenda/mknixboot (based on /rc/bin/mkusbboot) will help you
create a bootable USB stick from your NIX tree or, presumably, from a
9atom root.

/usr/glenda/bin/rc/startinstall is what you run after booting the USB stick.

Hope this helps.


John

[erik quanstrom]

On Fri Jan 4 11:36:45 EST 2013, ara...@mgk.ro wrote:
> Downloading from quanstro.net is extremely slow and I was never able
> to fully complete the download.

it's only a 384kbps line. ftp has some odd issue that is not
obvious from the code, but http download should work fine.
i'll also try to find the time to upgrade that machine with
the latest kernel, and thus the new tcp later today.

- erik

[erik quanstrom]

yes, the problems are mostly in the bios setup and details of
the hardware. (which may or may not be on-die or on package
with the cpu.)

i don't believe all the problems have been fixed. it certainly
was impossible to boot most atom machines at one point.

- erik

[David du Colombier]

> Downloading from quanstro.net is extremely slow and I was never able
> to fully complete the download.

A mirror is available on http://mirror.9grid.fr/

--
David du Colombier

[David du Colombier]

> There exist bootable images for USB sticks which contain a full fossil
> filesystem; you can start the installation from there.

A Plan 9 USB image with Fossil file system is available here:

http://www.9legacy.org/download/plan9-usb.img.bz2

It contains the full file system from the Plan 9 CD image,
unmodified, except a single line change in the kernel binary.

This image is a little old (2012-08-12), but I will probably
upload a newer one soon.

The instructions to build this image are available here:

http://www.9legacy.org/9legacy/doc/fossil/usb

It is very close to the mkusbboot(8) script.

--
David du Colombier

[Richard Miller]

> It contains the full file system from the Plan 9 CD image,
> unmodified, except a single line change in the kernel binary.

I can't contain my curiosity: what's the line?

I once worked with a colleague who had a superstition that any
one-line change to production software was almost certain to
introduce a bug. (Actually what he said was "one-card change",
which indicates how long ago this was.) If he had to make
a modification simple enough to be done in one line, he would
always combine it with some other random harmless change like
renaming a variable.

[David du Colombier]

> I can't contain my curiosity: what's the line?
>
> I once worked with a colleague who had a superstition that any
> one-line change to production software was almost certain to
> introduce a bug. (Actually what he said was "one-card change",
> which indicates how long ago this was.) If he had to make
> a modification simple enough to be done in one line, he would
> always combine it with some other random harmless change like
> renaming a variable.

This line:

http://www.9legacy.org/9legacy/patch/pc-sdata-dma-lite.diff

It enables DMA by default in the sdata driver.
But it's not a requirement. It's just a remain
of a personal taste.

Surprisingly, I have another version of this patch
which should convince your former colleague :-)

http://www.9legacy.org/9legacy/patch/pc-sdata-dma.diff

--
David du Colombier

[Jack Johnson]

It's funny, if you were correcting someone else's bug unknowingly introduced by a one-line change your fix could justifiably be one line, but then you would think the odds of it reoccurring in the future would be non-zero, and you might as well add a comment for future fumblers. :)

-Jack

[erik quanstrom]

> It's funny, if you were correcting someone else's bug unknowingly
> introduced by a one-line change your fix could justifiably be one line, but
> then you would think the odds of it reoccurring in the future would be
> non-zero, and you might as well add a comment for future fumblers. :)

insuring that asymptoticly, your code reaches heat death as the
comment/code ratio goes infinite. :-)

- erik

[erik quanstrom]

> Surprisingly, I have another version of this patch
> which should convince your former colleague :-)
>
> http://www.9legacy.org/9legacy/patch/pc-sdata-dma.diff

@@ -455,6 +455,8 @@
and
.I slot
to use as a root device for bootstrapping.
+.SS \fL*nodma=\fP
+Disable DMA on (S)ATA controllers.

it's a pretty consistent mistake in plan 9 (and elsewhere)
that ide is confused with ata or sata. these the are all
on different levels. "Disable DMA on IDE controllers."

ide device register set (analogue FIS)
sata on-the-wire protocol (analogue pata)
ata abstract command set.

ahci is not in this list because it's a programming
interface that is not directly tied to the hardware.
there are other programming interfaces for the
same hardware. (e.g. marvell & lsi interfaces.)

"sdata" is a misleading name, since it implies that it
actually handles the read/write ↔ ata command
translation. what it does is translate between
read/write and ata+ide in one fell swoop.

this sounds economical, until one realizes that things
like PUIS (power-up in standby) and SCT are fancy
protocols that happen at the ATA layer, and if not
abstracted, need to be repeated for every driver.

in essence, this is the rationale for the fis library:
it's tiresome rewriting the PUIS protocol.

- erik

[Rox 64]

I have that same netbook. One of your big challenges will probably be

getting the CDROM to boot in a machine without a CDROM drive. There
exist bootable images for USB sticks which contain a full fossil
filesystem; you can start the installation from there. In fact, I had
made a 32-bit bootable USB stick for Nix about a year ago and included
a script to start the installation, I'll see if I can find it.

I see. Thank you guys, been waiting to try Plan 9 since a while, but I though my hardware was incapable of booting and running it. In fact I wanted to buy a Raspberry Pi to install Miller's build... Although I will get sooner or later a Rpi haha.

What file system do I need to install Plan 9? Fat32? NTFS? Ext2/3/4?

[Matthew Veety]

Fossil, cwfs, or hjfs. 

[David du Colombier]

> it's a pretty consistent mistake in plan 9 (and elsewhere)
> that ide is confused with ata or sata. these the are all
> on different levels. "Disable DMA on IDE controllers."

Fixed. Thanks for the reminder.

--
David du Colombier

[erik quanstrom]

perhaps i'm being literal when it's not called for, but i don't
see this answers the question.

to *install* plan 9, you should be fine with a bare machine
plus the sd card or iso.

- erik

[Rox 64]

Yeah, my error was to assume I needed to make an special filesystem to boot and install Plan 9 from a USB device... Nevermind.

Thank you for the help guys, I'm going to try David's build. Wish me luck!

[Bakul Shah]

On Fri, 04 Jan 2013 14:44:43 EST erik quanstrom <quan...@quanstro.net> wrote:
>
> it's a pretty consistent mistake in plan 9 (and elsewhere)=20

> that ide is confused with ata or sata. these the are all
> on different levels. "Disable DMA on IDE controllers."
>
> ide device register set (analogue FIS)
> sata on-the-wire protocol (analogue pata)
> ata abstract command set.

Wikipedia says:

The terms "integrated drive electronics" (IDE), "enhanced
IDE" and "EIDE" have come to be used interchangeably with
ATA (now Parallel ATA, or PATA).

This matches my understanding. Recall that IDE was essentially
ST-506 with all the analog bits done on the drive itself
(hence the name Integrated Drive Electronics).

[erik quanstrom]

> Wikipedia says:
>
> The terms "integrated drive electronics" (IDE), "enhanced
> IDE" and "EIDE" have come to be used interchangeably with
> ATA (now Parallel ATA, or PATA).

i don't think so. these are the last set of ata8 docs i downloaded
from incits. there's no parallel spec at all. it's been dropped

aam architecture model
ast serial transport
acs ata command set

i don't see any room here for conflating ide with ata.

there's a seperate set of docs detailing sata from sata-io.

> This matches my understanding. Recall that IDE was essentially
> ST-506 with all the analog bits done on the drive itself
> (hence the name Integrated Drive Electronics).

just because it once was, so it shall always be? somehow
i think the world has moved on.

- erik

[Bakul Shah]

On Fri, 04 Jan 2013 15:36:06 EST erik quanstrom <quan...@quanstro.net> wrote:
> > This matches my understanding. Recall that IDE was essentially
> > ST-506 with all the analog bits done on the drive itself
> > (hence the name Integrated Drive Electronics).
>
> just because it once was, so it shall always be? somehow
> i think the world has moved on.

My only point was that there is no need to throw away history
and redefine commonly understood terms -- that only increases
confusion. New revisions of the ATA spec should use new terms
If old ones are found wanting but not redefine them.

[Matthew Veety]

Well he asked for the filesystem. I gave him the filesystems.
>

[David du Colombier]

> Thank you for the help guys, I'm going to try David's build.
> Wish me luck!

By the way, I have just uploaded a new USB image, based
on the latest Plan 9 CD image (2013-01-03).

--
David du Colombier

[erik quanstrom]

unfortunately, they have been redefined for us. ATA means
a command set. IDE is a programming interface. if that
has changed in the last 25 years, it is not i who changed the
definitions.

- erik

[Rox 64]

Hi. I have tried David's build. I copied it to an USB with ''dd if=[.img file] of=/dev/sdb bs=1M" and it boots. It detects the CPU and I can see something like "boot into real mode". Then nothing happens. The LED of my USB also does not indicate disk activity too.

What can I do now?

[Rox 64]

Ah, nevermind, I have seen that rc script, I was very stupid thinking I could do the same thing you can do with Miller's SD image for the Rpi. Please forgive my stupidity and (serious) brain damage.

[David du Colombier]

> Ah, nevermind, I have seen that rc script, I was very stupid thinking
> I could do the same thing you can do with Miller's SD image for the
> Rpi. Please forgive my stupidity and (serious) brain damage.

No, you did it right. The problem you reported is well-known
and happens with 9loadusb on some motherboards.
Unfortunately, there is still no known workaround.

--
David du Colombier

[Rox 64]

Maybe should I try 9front instead? I read they made a new bootloader to fix that issue.

[John Floren]

On Sat, Jan 5, 2013 at 2:42 PM, Rox 64 <mro...@gmail.com> wrote:
> Maybe should I try 9front instead? I read they made a new bootloader to fix
> that issue.

The PC kernel is also (supposed to be) multiboot-compliant, so you
should be able to boot it with GRUB if that helps. You just have to
build the kernel as an ELF file.

This means, of course, that you'll have to make do with default
configuration options or hard-code the options, since GRUB doesn't
read plan9.ini.


john

[Rox 64]

I don't really want to deal with ELF binaries and Grub...
I will try 9front and if its bootloader works then I will use it against vanilla Plan 9.

[Jacob Todd]

I've never had to do that when when using grub and plan 9,
chainloading always worked.

[John Floren]

I merely suggest GRUB because it seems like most of the time, if I can
get through 9load, I can boot the system OK, but sometimes 9load won't
work on such-and-such hardware and I have to resort to PXE booting,
putting the disk in another system to install, whatever. I've found
GRUB on a bootable USB stick to be a pretty reliable way to drop you
into a 32-bit kernel.

[Rox 64]

Ah, sorry then if my previous message sounded somehow violent.
My idea is to install and run Plan 9 on a USB a few days, then copy the disk image onto the HDD when I will be comfortable with it, so what I want is to be able to boot from USB every time I plug it without relying on external software such as GRUB and with hardcoded options.
But if I will take your suggestion if any other option fails =)

[Rox 64]

Sorry, I have another question here. I have browsed the wiki, the mailing list and I have googled it a lot but I'm still not sure.

9front's wiki says that in order to make a Live USB you need to copy a compiled Plan 9 kernel in a  '/386' folder. From what I have seen there's two available 386 kernels in 9front: 9pcf and 9pccpuf. If I'm correct, 9pcf is the standard kernel and 9pccpuf is for configuring a standalone CPU server, right?

[Matthew Veety]

Correct.

[Rox 64]

I see.

9front's booloader works flawlessly. It's a little annoying to write local!/shr/sdU7.0/tmp/9front.iso at the command prompt everytime, but at least it runs. The only problem (aside of using the US QWERTY layout, I'm spanish) is that it cannot run rio (it says something like 'cannot open /dev/draw/new: no frame buffer', I will investigate why it is happening). But now I can use 9front, and I'm happy with that. Now I will try running Plan 9 with 9front's FAT-based bootloader and see what happens.

Thank you for all the help, guys.

[erik quanstrom]

vgasize=XxYx16 # replace X and Y with something reasonable for your monitor.
monitor=vesa
aux/realemu
aux/vga -l $vgasize
rio

use kbmap(3) to fix your keyboard.

- erik

[Rox 64]

It seems I don't need aux/realemu and aux/vga at all to boot rio, it can run with monitor=vesa and vgasize=640x480x8 in plan9.ini. I still haven't found out how to run in 1024x738 with vesa, through, will check if it can run at least at 800x600 with the vga command as you said.

BTW, there are two things I have detected in Plan 9 that I love: the ability to copy and delete text in rio like if it were a simple file (much like the rest of Plan 9) and no CapsLock support (why other operating systems still support CapsLock? Seriously, it should die).

[John Floren]

On Sun, Jan 6, 2013 at 8:49 AM, Rox 64 <mro...@gmail.com> wrote:
> It seems I don't need aux/realemu and aux/vga at all to boot rio, it can run
> with monitor=vesa and vgasize=640x480x8 in plan9.ini. I still haven't found
> out how to run in 1024x738 with vesa, through, will check if it can run at
> least at 800x600 with the vga command as you said.

aux/vga and aux/mouse are being called by the startup scripts in that
case. Erik just posted the variable declarations and command calls
which end up being made from plan9.ini and the startup scripts--you'll
only need to use them if you've misconfigured plan9.ini.

john